When faults occur in microservice applications – as they inevitably do – developers depend on observability data to quickly identify and diagnose the issue. To collect such data, microservices need to be instrumented and the respective infrastructure configured. This task is often underestimated and error-prone, typically relying on many ad-hoc decisions. However, some of these decisions can significantly affect how quickly faults are detected and also impact the cost and performance of the application.
Given its importance, we emphasize the need for a method to guide the observability design process. In this paper, we build on previous work and integrate our observability experiment tool OXN into a novel method for continuous observability assurance. We demonstrate its use and discuss future directions.
